Terminal-Bench 2.0

Terminal-Bench~2.0 is a benchmark dataset used to evaluate the performance of large language models as agents in long-horizon tasks by assessing their interaction with environments through various harness configurations.

Papers using Terminal-Bench 2.0 (9)

Endless Terminals: Scaling RL Environments for Terminal Agents2026On Data Engineering for Scaling LLM Terminal Capabilities2026ECHO: Terminal Agents Learn World Models for Free2026HarnessBridge: Learnable Bidirectional Controller for LLM Agent Harness2026SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolution2026LiteCoder-Terminal: Scaling Long-Horizon Terminal Environments for Learning Language Agents2026Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills2026Tmax: A simple recipe for terminal agents2026How Well Do Agentic Skills Work in the Wild: Benchmarking LLM Skill Usage in Realistic Settings2026